Table 1.
Publications relevant to RT prediction.
| Publication | Year | LC type | Model type | Size of training data | Molecular type | Variables |
|---|---|---|---|---|---|---|
| Hagiwara et al. [175] | 2010 | RP-LC | SVR and MLR | 150 authentic compounds | 9 MDs | |
| Creek et al. [176] | 2011 | HILIC | MLR | 120 authentic compounds | 6 MDs | |
| D'Archivio, Maggi and Ruggieri [177] | 2014 | RP-LC | MLR and PLS regression | 47 authentic compounds | butyl esters of 47 acylcarnitines | 73 MDs |
| Kouskoura, Hadjipavlou-Litina and Markopoulou [178] | 2014 | RP-LC | PLS regression | 100 authentic compounds | 66 MDs | |
| D'Archivio et al. [179] | 2014 | RP-LC | DNNs | 24 authentic compounds | s-triazines | 5 MDs |
| Cao et al. [180] | 2015 | HILIC | MLR and RF | 93 authentic compounds | 346 MDs | |
| Aicheler et al. [181] | 2015 | RP-LC | SVR | 201 authentic compounds | lipid | 11 MDs |
| Munro et al. [182] | 2015 | RP-LC | DNNs | 166 authentic compounds | drugs | 17 MDs |
| Falchi et al. [183] | 2016 | RP-LC | Four combined (fingerprints + ordinary) KPLS models | 1383 authentic compounds | molecular and fingerprints descriptors | |
| Ovcacikova et al. [184] | 2016 | RP-LC | The second degree polynomial regression | 400 authentic compounds | lipid | The carbon number (CN) and the double bonds (DB) number |
| Aalizadeh et al. [185] | 2016 | RP-LC | MLR, DNNs, and SVM | 528 and 298 compounds for positive and negative electrospray ionization mode respectively | 6 MDs | |
| Wolfer et al. [186] | 2016 | RP-LC | Combination of RF and SVR models | 442 authentic compounds | 97 MDs | |
| Kubik and Wiczling [187] | 2016 | RP-LC | Lasso, Stepwise and PLS regressions | 115 authentic compounds | drugs | 50 MDs |
| Barron and McEneff [188] | 2016 | RP-LC | DNNs | 1,117 authentic compounds | 16 MDs | |
| Randazzo et al. [189] | 2016 | RP-LC | PLS regression | 91 authentic compounds | steroids | 97 MDs |
| Taraji et al. [190] | 2017 | HILIC | PLS regression | 16 authentic compounds | β-adrenergic agonists and related compounds | 321 MDs |
| Taraji et al. [191] | 2017 | HILIC | PLS regression | 98 authentic compounds | pharmaceutical compounds | 321 MDs |
| Zhang et al. [192] | 2017 | RP-LC | MLR | 24 authentic compounds | 16-membered ring macrolides | 8 MDs |
| Park et al. [193] | 2017 | RP-LC | MLR | 41 authentic compounds | drugs | 10 MDs |
| Wen et al. [194] | 2018 | RP-LC | PLS regression | 148 authentic compounds | 126 MDs | |
| Wen et al. [195] | 2018 | RP-LC | PLS regression | 191 authentic compounds | 128 MDs | |
| McEachran et al. [196] | 2018 | RP-LC | PLS regression | 97 authentic compounds | 7 MDs | |
| Hall et al. [197] | 2018 | RP-LC | DNNs | 1,955 authentic compounds | 47 MDs | |
| Bouwmeester, Martens and Degroeve [198] | 2019 | RPLC (33) & HILIC (3) | Bayesian Ridge Regression (BRR), Least Absolute Shrinkage and Selection Operator (LASSO), DNNs, Adaptive Boosting (AB), Gradient Boosting (GB), RF and SVR | 6,759 authentic compounds | 151 MDs | |
| Bonini et al. [154] | 2020 | HILIC & RP-LC | XGBoost, Bayesian-regularized Neural Network (BRNN), RF, Light Gradient-Boosting Machine (LightGBM), DNNs | 1,023 (HILIC) & 494 (RP-LC) authentic compounds | 286 MDs | |
| Ju et al. [163] | 2021 | HILIC & RP-LC | DNNs + TL | 77,898 authentic compounds (DNNs), and 17 data sets (Transfer Learning) | 1,470 MDs | |
| Osipenko et al. [159] | 2021 | HILIC & RP-LC | RNNs + TL | 1 million molecules (pre-training) and 269–457 authentic compounds (transfer Learning) | SMILES | |
| Kensert et al. [156] | 2021 | HILIC & RP-LC | Graph Convolutional Networks (GCNs) | 77,980 (SMRT), 852(RIKEN) and 1,400 (Fiehn HILIC) authentic molecules | Graph and 25 atom and bond features | |
| Yang et al. [157] | 2021 | HILIC | GNNs + TL | in silico HILIC RT dataset with about 306 K molecules for GNNs, 100∼200 molecules for TL | Graph, 16 kinds of atoms and 4 kinds of bonds | |
| Yang et al. [158] | 2021 | RP-LC | GNNs + TL | 80,038 authentic molecules (SMRT) for Graph Neural Network, and the MoNA and PredRet datasets for Transfer Learning | Graph | |
| Souihi et al. [199] | 2022 | HILIC & RP-LC | RF regression | 78 authentic compounds | 153 MDs | |
| Liapikos et al. [200] | 2022 | RP-LC | Bayesian Ridge Regression (BRidgeR), Extreme Gradient Boosting Regression (XGBR) and SVR | 26–350 authentic compounds | 70–92 MDs | |
| Fedorova et al. [155] | 2022 | RP-LC | 1D CNN + TL | 77,983 authentic molecules (SMRT) for 1D CNN, 5 data sets for Transfer Learning | SMILES |